Intelligent delay closing electric energy meter control platform and method

Abstract

The invention belongs to the technical field of electric energy meters, and discloses an intelligent delay closing electric energy meter control platform and method. The method comprises the following steps: acquiring power consumption data and power consumption environment data of an electric energy meter; constructing a multi-dimensional data space based on the power utilization data and the power utilization environment data of the electric energy meter; constructing a two-dimensional grid evolution model based on the multi-dimensional data space; identifying a power consumption peak region in the two-dimensional grid evolution model; for the identified power utilization peak value region, an improved sphere optimization algorithm is adopted to determine the optimal delay closing time and duration so as to achieve the purpose of peak clipping and valley filling; the operation of the power grid is optimized, and better power utilization service is provided for users.

Description

Technical Field

[0001] The present invention relates to the technical field of electric energy meters, and more specifically, to an intelligent delayed closing electric energy meter control platform and method. Background Art

[0002] With the rapid development and popularization of smart grids, the control technology of electric energy meters, as key electricity consumption data collection equipment, is also constantly updated; traditional electric energy meter control systems mainly rely on simple electricity consumption records and basic load management, but they seem to be unable to cope with the increasingly complex electricity consumption environment and diversified user needs.

[0003] Traditional control platforms find it difficult to effectively process and analyze massive amounts of multi-dimensional electricity consumption data, making it impossible to fully explore the deep electricity consumption patterns and laws contained in the data; this makes it difficult to accurately predict peak electricity consumption and to take effective load management measures in a timely manner; secondly, when identifying peak electricity consumption areas, existing methods are often too simplistic and static, and cannot accurately capture dynamically changing electricity consumption behaviors, especially electricity consumption fluctuations during holidays or special weather conditions; in addition, traditional delayed closing strategies lack flexibility and intelligence, and are difficult to accurately adjust based on real-time electricity consumption and historical data, often resulting in unnecessary power outages or missing critical peak-shaving opportunities; more importantly, existing technologies often focus too much on short-term effects when implementing peak-shaving and valley-filling, and ignore the comprehensive consideration of multiple factors such as user experience and long-term impacts; this one-sided control strategy has a negative impact on certain important power facilities and affects overall stability and reliability.

[0004] In view of this, the present invention proposes an intelligent delayed closing electric energy meter control platform and method to solve the above problems. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solutions: an intelligent time-delay closing electric energy meter control platform, comprising: a data acquisition module, for acquiring the electric energy meter's power consumption data and power consumption environment data; The space construction module constructs a multi-dimensional data space based on the electricity consumption data and electricity consumption environment data of the energy meter; The model building module builds a two-dimensional grid evolution model based on the multi-dimensional data space; An identification module is used to identify the peak power consumption area in the two-dimensional grid evolution model; The optimization control module is used to determine the optimal delay closing time and duration for the identified power consumption peak area by using an improved sphere optimization algorithm; each module is connected by wired and / or wireless means.

[0006] Furthermore, the electricity consumption data includes reading data, electricity power curve, electricity load type, electricity equipment type and user type; the electricity consumption environment data includes meteorological data, holiday information, electricity consumption area information and grid load level.

[0007] Furthermore, the method of constructing the multidimensional data space includes: The electricity consumption data and electricity consumption environment data of the electric energy meter are cleaned and standardized, and the categorical data are numerically encoded to obtain the preprocessed electricity consumption data and electricity consumption environment data; the preprocessed electricity consumption data and electricity consumption environment data are merged into a data set X; each row in the data set represents a sample; the sample is constructed into a vector representation, which is recorded as a sample vector; For each sample x in the data set X, calculate the distance between it and all other samples; based on the obtained distance, construct an n×n distance matrix D, where n is the number of samples in the data set X, and the element D(i, j) in the distance matrix D represents the distance between the i-th sample and the j-th sample; For each sample x, find the k samples closest to x in the distance matrix D, that is, the k nearest neighbors of x; k is a preset value and an integer greater than 1; the sample x and its k nearest neighbors form a set, recorded as the nearest neighbor set N(x); construct an undirected weighted graph G=(V, E, W), where V is the set of all nodes in the undirected weighted graph, corresponding to all samples in the data set X; E is the set of all edges in the undirected weighted graph; W is the set of edge weights; convert the undirected weighted graph G into a fuzzy geometric graph; optimize the fuzzy geometric graph for dimensionality reduction to obtain a multidimensional data space.

[0008] Furthermore, the method of finding k samples closest to x in the distance matrix D includes: For each sample x, initialize k empty lists, copy the row corresponding to sample x in the distance matrix D to a temporary list T1, sort the temporary list T1 to obtain an ordered list in increasing distance, and divide the ordered list into k sublists of equal size; Take out the element with the smallest distance from each sublist to form a minimum heap, take out the element with the smallest distance from the minimum heap, add it to the ordered list S, insert the next element in the sublist where the taken element is located into the minimum heap, and repeat until all sublists are merged into the ordered list S. In the ordered list S, take out the first k elements, and their corresponding samples are the k nearest neighbors of x.

[0009] Furthermore, the method of converting the undirected weighted graph G into a fuzzy geometric graph includes: For each node in the undirected weighted graph G , calculate its comprehensive local density ; ; in, For nodes The nearest neighbor set The nodes inside For Node and nodes The distance between is a control parameter, and its value is in the range of (0, 1); for To its The distance to the nearest neighbor, For Node and nodes The importance factor between Importance Factor ;in, For Node and nodes The similarity score between For Node and nodes The connectivity score between is the weight coefficient, and its value is in the range of (0, 1); ;in, For Node The corresponding sample vector is For Node The corresponding sample vector is It is the vector dot product operation; If the node and nodes There exists a shortest path in an undirected weighted graph G ;but ; If the node and nodes In the undirected weighted graph G, they are directly connected, then ; Find the ratio The comprehensive local density All nodes with large size form a set, which is recorded as the large density set ,for Each node in ,calculate and The distance between ,Pick All nodes in The minimum distance ; For each node , calculate a clustering factor , sort all nodes in descending order according to the value of the clustering factor, start from the node with the largest clustering factor value, divide it and its nearest neighbors into a cluster, then move to the next undivided node with the largest clustering factor value, and repeat until all nodes are divided into clusters; For each cluster, it is regarded as a simple geometry; for each node belonging to the cluster, its membership is defined as the inverse of the distance between the corresponding node and the central node of the corresponding cluster; the combination of the membership of all simple geometries and their corresponding nodes constitutes a fuzzy geometric graph Gg.

[0010] Furthermore, the method of performing the optimization and dimensionality reduction process includes: Define a low-dimensional space and determine the target dimension d of the low-dimensional space; for each sample x in the data set X, take it as a data point and project it into the d-dimensional space to obtain a d-dimensional vector as the initial position of the data point in the low-dimensional space; in the low-dimensional space, based on the initial position of the data point, construct a low-dimensional fuzzy simple geometric graph G_low; Define the loss function ; ; in x belongs to the convex hull shape in the fuzzy geometric graph Gg The membership degree of is the degree of membership of x in the low-dimensional fuzzy simple geometric graph G_low to the convex hull shape s; is the weight coefficient of the regularization term, is the neighbor distribution of x in the fuzzy geometric graph Gg, is the neighbor distribution of x in the low-dimensional fuzzy simple geometric graph G_low; Represents KL divergence; based on the loss function, the position of the data points in the low-dimensional space is iteratively optimized to minimize the loss function. In each iterative optimization, based on the position of the current data point, a low-dimensional fuzzy simple geometric graph is constructed, the value of the current loss function is calculated, and the position of each data point is back-propagated to obtain the gradient, and the optimizer is used to update the position of each data point according to the gradient; repeat until the preset maximum number of iterations is reached; the final low-dimensional fuzzy simple geometric graph is the multidimensional data space.

[0011] Furthermore, the two-dimensional grid evolution model is constructed in the following manner: The multidimensional data space is mapped onto a two-dimensional plane, and an N×N square grid is constructed on the mapped two-dimensional plane. Each grid point in the square grid represents a local area in the multidimensional data space. Each grid point is randomly assigned an initial state, which is either alive or dead. For each grid point, its neighbors are defined, and the neighbors include 8 adjacent grid points in the up, down, left, right, and four diagonal directions. For the square grid, set the evolution rule. The evolution rule is: if a grid point in the live state has 2 to 3 neighbors in the live state, it will remain in the live state in the next generation. If a grid point in the live state has less than 2 or more than 3 neighbors in the live state, it will become dead in the next generation. If a grid point in the dead state has exactly 3 neighbors in the live state, it will become alive in the next generation. Define the length of the time step; starting from the initial state, iteratively update the states of all grid points according to the evolution rules. In each iteration, for each grid point, find its neighbors in the grid and calculate their states. According to the evolution rules, determine the state of the corresponding grid point in the next time step, and update the states of all grid points at the same time; and record the number of grid points whose states are alive at each time step. The grid whose states are iteratively updated is the two-dimensional grid evolution model.

[0012] Furthermore, the method for identifying the peak power consumption area includes: Set a time window of fixed length, slide the time window along the time axis, and move one time step each time; in the process of sliding the time window, calculate the activity of each grid point in the time window; the calculation formula of activity is: ;in, is the number of time steps in the time window, is the index of the time step, Grid In time step status, For the time step Time Grid The number of live neighbors among the surrounding 8 neighbors, is the neighborhood influence coefficient, and its value is in the range of (0, 1); Grid activity level; is the step size control parameter; The activity of the grid points in the time window is organized into a matrix, recorded as the activity matrix, and Gaussian smoothing or mean filtering is applied to the activity matrix; a high activity threshold is defined, the activity matrix is ​​traversed, and the continuous area formed by the grid points whose activity is higher than the high activity threshold is marked as a potential peak area; Convert the activity matrix into a binary image, and set the pixels corresponding to the potential peak area on the binary image to 1, and the others to 0; obtain a preliminary binary image; apply dilation and erosion operations to the preliminary binary image; obtain the potential peak area in the middle section; In the process of sliding the time window, for adjacent time windows, calculate the overlap of the potential peak areas in the middle section obtained by them; define an overlap threshold; if the overlap between the current time window and a previous time window is higher than the overlap threshold, then record the duration between the two time windows; A minimum duration threshold mdu is set, and those mid-segment potential peak areas whose duration exceeds mdu are marked as stable areas. The tracking period and average threshold are set. For each pair of stable areas, their average overlap during the entire tracking period is calculated. If the average overlap is higher than the average threshold, the corresponding two stable areas are merged, that is, the union of the two corresponding stable areas in all time windows is calculated, and this union is used as the new merged area. The new merged area is morphologically processed to obtain the final power consumption peak area.

[0013] Furthermore, the optimal delay closing time and duration are determined by: Define decision variables, including the delay closing time and duration ; Define the objective function ; ; in , , as well as is the target weight coefficient, for The original power consumption at the time, To implement delayed closing The power consumption at the time, represents the positive part, is the cumulative effect factor of historical delayed impact; is the user impact factor, is the load balancing factor; ;in, is the power change rate weight coefficient; is the historical impact weight coefficient; for The power change at each moment, is the time interval; is a delayed time point in history; is the time decay constant; ;in, For the The importance weight coefficient of the class user, for Moment The number of users of the class; For the Class User The accumulated delay time of the moment; ;in, is the load balance factor; is the maximum load value during the observation period, To implement delayed closing The actual load value at the moment, is the minimum load value during the observation period; Randomly generate N2 spheres, where N2 is an integer greater than 1. Each sphere contains a center position and a radius. The center position is composed of decision variables, i.e. (x1, x2). All spheres are sorted in descending order according to the function value of the objective function. For each sphere, randomly generate M candidate solutions on the surface of the sphere, calculate the function value of the objective function of the candidate solution, take the candidate solution with the largest function value as the optimal candidate solution, update the center position of the sphere to the position of the optimal candidate solution, and adjust the radius of the sphere. The formula for adjusting the radius of the sphere is: ;in, is the adjusted radius, is the sphere corresponding to the optimal candidate solution The radius of The sphere selected from all spheres The sphere with the largest function value of the candidate solution ; is the random disturbance factor, and its value range is (-0.1, 0.1); All spheres are sorted, spheres with smaller function values ​​are eliminated, and spheres with adjusted radius are used as new spheres; the change rate of the optimal candidate solution is iteratively and repeatedly calculated. If the change rate is less than the preset change threshold for K consecutive times, the last optimal candidate solution is used as the final solution, and the decision variables corresponding to the final solution are extracted to obtain the optimal delayed closing time and duration.

[0014] A method for controlling an intelligent time-delay closing electric energy meter is implemented based on the intelligent time-delay closing electric energy meter control platform, comprising: step 1, obtaining power consumption data and power consumption environment data of the electric energy meter; Step 2: construct a multidimensional data space based on the electricity consumption data and electricity consumption environment data of the energy meter; Step 3: Based on the multidimensional data space, a two-dimensional grid evolution model is constructed; Step 4, identifying the peak power consumption area in the two-dimensional grid evolution model; Step 5: For the identified peak power consumption area, the improved sphere optimization algorithm is used to determine the optimal delayed closing time and duration.

[0015] The technical effects and advantages of the intelligent time-delay closing electric energy meter control platform and method of the present invention are as follows: The present invention can effectively improve the operating efficiency and stability of the power grid and realize accurate power load management; by deeply mining the complex power consumption data, it can accurately identify the dynamically changing power consumption peak area, thereby realizing more flexible and accurate peak shaving and valley filling control; this not only helps to balance the power grid load, but also can significantly reduce the operating cost of the power system; at the same time, it fully considers the user experience, and can formulate more humane control strategies according to different user types and historical power consumption conditions, effectively reducing the impact on users' daily power consumption; in addition, the intelligent characteristics enable it to adapt to the complex and changeable power consumption environment, improve the adaptability and reliability of power management, and generally speaking, it not only optimizes the power grid operation, but also provides users with better power consumption services. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic diagram of an intelligent time-delay closing electric energy meter control platform of the present invention; Figure 2 The present invention is a schematic diagram of an intelligent time-delay closing electric energy meter control method. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] Example 1 See also Figure 1 As shown, the present embodiment provides an intelligent time-delay closing electric energy meter control platform, including: a data acquisition module, used to obtain the electric energy meter's power consumption data and power consumption environment data; The space construction module constructs a multi-dimensional data space based on the electricity consumption data and electricity consumption environment data of the energy meter; The model building module builds a two-dimensional grid evolution model based on the multi-dimensional data space; An identification module is used to identify the peak power consumption area in the two-dimensional grid evolution model; The optimization control module is used to determine the optimal delayed closing time and duration for the identified peak power consumption area using an improved sphere optimization algorithm to achieve the purpose of peak shaving and valley filling; each module is connected by wired and / or wireless means to achieve data transmission between modules.

[0019] The electricity consumption data includes reading data, electricity power curve, electricity load type (numerical code), electricity equipment type (numerical code) and user type (numerical code); the reading data includes total electricity consumption (kWh) and electricity consumption by time period (kWh / time period); electricity power (kilowatt / time point), reflecting the change of electricity power over time; electricity load type, such as residential electricity, industrial electricity, commercial electricity, etc.; electricity equipment type, such as air conditioning, lighting, electrical appliances, etc.; user type, such as household users, corporate users, etc.; the reading data and electricity power curve are sequence data, which change with time. For the type of data, the numerical code is used, such as 1 for residential electricity and 2 for industrial electricity.

[0020] Electricity consumption environment data includes meteorological data, holiday information (0 / 1 coding), electricity consumption area information (numerical coding) and grid load level (percentage); meteorological data includes temperature, humidity and sunshine duration, environmental factors that affect electricity consumption; electricity consumption area information, such as city, region, etc.

[0021] Directly processing the original high-dimensional data will face the dimensionality curse, high computational complexity, low efficiency, and simple dimensionality reduction methods (such as principal component analysis) may lose important nonlinear relationships and local structural information. In the original high-dimensional space, the distribution and relationship of the data may not be obvious, making it difficult to discover meaningful electricity consumption patterns and laws.

[0022] Ways to construct a multidimensional data space include: Clean and standardize the electricity consumption data and electricity environment data of the electric energy meter, and handle missing values, abnormal values, etc.; perform numerical coding on categorical data (such as electricity load type, electricity equipment type, user type, electricity area information, etc.) to obtain pre-processed electricity consumption data and electricity environment data; The preprocessed electricity consumption data and electricity consumption environment data are combined into a data set X; each row in the data set represents a sample, including all electricity consumption data (such as reading data, electricity power curve, etc.) and electricity consumption environment data (such as meteorological data, holiday information, etc.); the sample is constructed into a vector representation, recorded as a sample vector.

[0023] For each sample x in the data set X, calculate the distance between it and all other samples. The distance is calculated by the Euclidean distance, Manhattan distance or cosine similarity between the sample vectors corresponding to the sample; Based on the obtained distance, an n×n distance matrix D is constructed, where n is the number of samples in the data set X, and the element D(i, j) in the distance matrix D represents the distance between the i-th sample and the j-th sample.

[0024] For each sample x, find the k samples closest to x in the distance matrix D, that is, the k nearest neighbors of x; k is a preset value and an integer greater than 1; specifically, for each sample x, initialize k empty lists, copy the row corresponding to sample x in the distance matrix D (indicating the distance between sample x and all other samples) to a temporary list T1, sort the temporary list T1 to obtain an ordered list in increasing distance, and divide the ordered list into k sublists of equal size.

[0025] Take out the element with the smallest distance from each sublist to form a minimum heap, take out the element with the smallest distance from the minimum heap, add it to the ordered list S, insert the next element in the sublist where the taken element is located into the minimum heap, and repeat until all sublists are merged into the ordered list S. In the ordered list S, take out the first k elements, and their corresponding samples are the k nearest neighbors of x; the sample x and its k nearest neighbors form a set, recorded as the nearest neighbor set N(x).

[0026] Construct an undirected weighted graph G=(V, E, W), where V is the set of all nodes in the undirected weighted graph, corresponding to all samples (nodes) in the data set X; E is the set of all edges in the undirected weighted graph, that is, if sample x and sample y are each other's nearest neighbors, then add an undirected edge l(x, y) between x and y; W is the set of edge weights, where the element W(x, y) represents the weight of edge l(x, y), which is set to the distance between x and y; convert the undirected weighted graph G into a fuzzy geometric graph.

[0027] Specifically, for each node in the undirected weighted graph G , calculate its comprehensive local density ; ; in, For nodes The nearest neighbor set The nodes inside For Node and nodes The distance between (the similarity between sample vectors); is a control parameter, and its value is in the range of (0, 1); for To its Nearest neighbors (except distance outside For Node and nodes The importance factor between them.

[0028] Importance Factor ;in, For Node and nodes The similarity score between For Node and nodes The connectivity score between is the weight coefficient, which is used to balance the weights of the two items and its value is in the range of (0, 1).

[0029] ;in, For Node The corresponding sample vector is For Node The corresponding sample vector is It is a vector dot product operation; if the node and nodes There exists a shortest path in an undirected weighted graph G ;but ; If the node and nodes In the undirected weighted graph G, they are directly connected, then .

[0030] Find the ratio The comprehensive local density All nodes with large size form a set, which is recorded as the large density set ,for Each node in ,calculate and The distance between ,Pick All nodes in The minimum distance .

[0031] For each node , calculate a clustering factor , sort all nodes in descending order according to the value of the clustering factor, start from the node with the largest clustering factor value, divide it and its nearest neighbors into a cluster, then move to the next undivided node with the largest clustering factor value, and repeat until all nodes are divided into clusters.

[0032] For each cluster, it is regarded as a simple geometry; for each node belonging to the cluster, its membership is defined as the inverse of the distance between the corresponding node and the central node of the corresponding cluster; the combination of the membership of all simple geometries and their corresponding nodes constitutes a fuzzy geometric graph Gg; in which each data point is assigned a membership to a simple geometry; this representation retains the local and global topological structure of the data.

[0033] The fuzzy geometric graph is optimized and reduced in dimension. Specifically, a low-dimensional space is defined and the target dimension d of the low-dimensional space is determined, usually d=2 or 3, for visualization. For each sample x in the data set X, it is taken as a data point and projected into the d-dimensional space to obtain a d-dimensional vector as the initial position of the data point in the low-dimensional space. The projection can be a random projection method.

[0034] In low-dimensional space, based on the initial position of the data point, a low-dimensional fuzzy simple geometric graph G_low is constructed. The calculation method is similar to the fuzzy geometric graph. The loss function is defined as ; ; in x belongs to the convex hull shape in the fuzzy geometric graph Gg The membership degree of is the degree of membership of x in the low-dimensional fuzzy simple geometric graph G_low to the convex hull shape s; is the weight coefficient of the regularization term, is the neighbor distribution of x in the fuzzy geometric graph Gg, is the neighbor distribution of x in the low-dimensional fuzzy simple geometric graph G_low; represents KL divergence; it needs to be explained that the convex hull It is a basic geometric structure in a geometric graph, used to represent the proximity and topological structure between data points. Specifically, a J-simplex is a convex hull consisting of J+1 linearly independent points (vertices); for example, a 0-simplex is a single point, a 1-simplex is a line segment consisting of 2 vertices, a 2-simplex is a triangle consisting of 3 vertices, a 3-simplex is a tetrahedron consisting of 4 vertices, and so on; the neighbor distribution is represented by a probability distribution, in which each value corresponds to the inverse of the distance from x to another node in the cluster.

[0035] Based on the loss function, the positions of data points in the low-dimensional space are iteratively optimized to minimize the loss function. In each iterative optimization, a low-dimensional fuzzy simple geometric graph is constructed based on the current position of the data point, the value of the current loss function is calculated, and the position of each data point is back-propagated to obtain the gradient. The optimizer (such as Adam) is used to update the position of each data point according to the gradient; repeat until the preset maximum number of iterations is reached; the final low-dimensional fuzzy simple geometric graph is the multidimensional data space.

[0036] By minimizing the cross entropy loss function, the low-dimensional fuzzy simple geometric graph in the low-dimensional space is forced to be as close as possible to G_high in the high-dimensional space; since G_high retains the local and global topological structure of the original data, the optimized low-dimensional fuzzy simple geometric graph will also retain these structural information as much as possible; specifically, if two samples are neighbors in the high-dimensional space (belonging to the same convex hull), then their positions in the low-dimensional space will also be brought closer; conversely, if they are not neighbors in the high-dimensional space, then the distance in the low-dimensional space will be pulled apart.

[0037] The electricity consumption data and electricity consumption environment data of the electric energy meter are projected into a confidentiality space, thus constructing a multidimensional data space; in this multidimensional data space, similar samples are close to each other, while different samples are far apart, which well preserves the essential structure and characteristics of the original high-dimensional data.

[0038] The construction methods of the two-dimensional grid evolution model include: The multidimensional data space is mapped onto a two-dimensional plane. For example, an algorithm such as t-SNE is used to construct an N×N square grid on the mapped two-dimensional plane. Each grid point in the square grid represents a local area in the multidimensional data space, and the distance between the grid points reflects the similarity of the data points in the original multidimensional data space. Each grid point is randomly assigned an initial state, which is either alive (1) or dead (0). For each grid point, its neighbors are defined, and the neighbors include 8 adjacent grid points in the top, bottom, left, right, and four diagonal directions. For the grid points at the boundary of the square grid, their opposite sides are connected to form a torus structure, which means that the left boundary of the grid is connected to the right boundary, and the upper boundary is connected to the lower boundary.

[0039] The evolution rules are set for the square grids. The evolution rules are: if a grid point in the live state is surrounded by 2 to 3 neighbors in the live state, it will remain in the live state in the next generation. If a grid point in the live state is surrounded by less than 2 or more than 3 neighbors in the live state, it will become dead in the next generation; if a grid point in the dead state is surrounded by exactly 3 live neighbors, it will become alive in the next generation.

[0040] Define the length of the time step and determine the actual length of time represented by each state update, such as 15 minutes or 1 hour; starting from the initial state, iteratively update the states of all grid points according to the evolution rules. In each iteration, for each grid point, find its neighbors in the square grid and calculate their states. According to the evolution rules, determine the state of the corresponding grid point in the next time step, and update the states of all grid points at the same time; and record the number of grid points with active states at each time step. The square grid that is iteratively updated in this state is the two-dimensional grid evolution model; the two-dimensional grid evolution model can better capture the complex relationships and structures of the original high-dimensional data, and can more accurately simulate and predict complex electricity consumption behavior patterns.

[0041] Peak electricity consumption areas can be identified by: Set a time window of fixed length, such as 4 hours or 8 hours, to observe the state changes of the two-dimensional grid evolution model, slide the time window along the time axis, and move one time step (such as 15 minutes or 1 hour) each time; in the process of sliding the time window, calculate the activity of each grid point in the time window; the calculation formula of activity is: ;in, is the number of time steps in the time window, is the index of the time step, Grid In time step The state (1 for alive, 0 for dead), For the time step Time Grid The number of live neighbors among the surrounding 8 neighbors, is the neighborhood influence coefficient, which is in the range of (0, 1) and is used to balance local and global information, that is, to strike a balance between considering the state of a single grid point and the state of its neighborhood; Grid activity level; is the step length control parameter, which takes a value in the range of (0, 1) and is used to adjust the balance between the distances between steps.

[0042] The activity of the grid points in the time window is organized into a matrix, recorded as the activity matrix, and Gaussian smoothing or mean filtering is applied to the activity matrix to reduce noise and highlight the main features; a high activity threshold (such as 0.8) is defined, the activity matrix is ​​traversed, and the continuous area composed of grid points with activity higher than the high activity threshold is marked as a potential peak area.

[0043] The activity matrix is ​​converted into a binary image, and the pixels corresponding to the potential peak area on the binary image are set to 1, and the others are set to 0; a preliminary binary image is obtained; dilation and erosion operations are applied to the preliminary binary image to merge similar areas and remove small isolated areas; the potential peak area in the middle section is obtained.

[0044] In the process of sliding the time window, for adjacent time windows, the overlap of the potential peak areas in the middle section is calculated; the overlap is measured by the Jaccard coefficient; an overlap threshold (such as 0.6) is defined to determine whether the area persists; if the overlap between the current time window and a previous time window is higher than the overlap threshold, it is considered to be a continuation of the same area; the duration between the two time windows is recorded; if it is not higher than the overlap threshold, it is regarded as a newly appeared area and the duration of each area is updated.

[0045] Set a minimum duration threshold mdu (such as 3 time windows), mark those mid-segment potential peak areas with a duration exceeding mdu as stable areas; set the tracking period and average threshold, and for each pair of stable areas, calculate their average overlap during the entire tracking period. If the average overlap is higher than the average threshold, merge the corresponding two stable areas, that is, calculate the union of the two corresponding stable areas in all time windows, use this union as the new merged area, perform morphological processing (such as dilation or closing operation) on the new merged area to smooth the boundaries and obtain the final power consumption peak area.

[0046] The best way to determine the delay closing time and duration is as follows: Define decision variables, including the delay closing time (minutes) and duration (minutes); define the objective function ; ; in , , as well as is the target weight coefficient, for The original power consumption at the time, To implement delayed closing The power consumption at the time, Indicates the positive part, that is, only the reduction amount is counted, is the cumulative effect factor of historical delayed impact; is the user impact factor, is the load balancing factor.

[0047] ;in, is the power change rate weight coefficient, which is used to control the importance of the power change speed, and its value range is (0, 1); is the historical impact weight coefficient, which is used to control the importance of the historical delay impact, and its value range is (0, 1); for The power change at each moment, is the time interval, usually 15 minutes or 1 hour; is a delayed time point in history; is the time decay constant, which controls the decay speed of historical influence, and the unit is hour.

[0048] ;in, For the The importance weight coefficient of a user type, with a value range of (0, 1); for example, important industrial users have a larger weight (0.8-1.0), while ordinary residential users have a smaller weight (0.3-0.5); for Moment The number of class users, integer value; For the Class User The cumulative delay time of the moment, in hours.

[0049] ;in, is the load balancing coefficient, which is used to adjust the importance of the valley filling effect and has a value range of (0, 1); is the maximum load value during the observation period, To implement delayed closing The actual load value at the moment, is the minimum load value during the observation period.

[0050] Randomly generate N2 spheres, where N2 is an integer greater than 1. Each sphere contains a center position and a radius. The center position is composed of decision variables, i.e. (x1, x2). All spheres are sorted in descending order according to the function value of the objective function. For each sphere, randomly generate M candidate solutions (center positions) on the surface of the sphere, calculate the function value of the objective function of the candidate solution, take the candidate solution with the largest function value as the optimal candidate solution, update the center position of the sphere to the position of the optimal candidate solution, and adjust the radius of the sphere. The formula for adjusting the radius of the sphere is: ;in, is the adjusted radius, is the sphere corresponding to the optimal candidate solution The radius of The sphere selected from all spheres The sphere with the largest function value of the candidate solution ; is a random perturbation factor with a value range of (-0.1, 0.1) to increase diversity and avoid having the same search range.

[0051] Sort all spheres, eliminate spheres with smaller function values, and use spheres with adjusted radius as new spheres; iteratively and repeatedly calculate the change rate of the optimal candidate solution, mainly considering the change of function value, the change of delayed closing time (x1) and the change of duration (x2); combine the changes of these three aspects in a weighted manner; if the change rate is less than the preset change threshold for K consecutive times, the last optimal candidate solution is taken as the final solution, and the decision variables corresponding to the final solution are extracted to obtain the optimal delayed closing time and duration.

[0052] This embodiment can effectively improve the operating efficiency and stability of the power grid and realize accurate power load management; by deeply mining complex power consumption data, it can accurately identify dynamically changing power consumption peak areas, thereby realizing more flexible and accurate peak shaving and valley filling control; this not only helps to balance the power grid load, but also can significantly reduce the operating cost of the power system; at the same time, it fully considers the user experience, and can formulate more humane control strategies according to different user types and historical power consumption conditions, effectively reducing the impact on users' daily power consumption; in addition, the intelligent characteristics enable it to adapt to complex and changeable power consumption environments, improve the adaptability and reliability of power management, and generally speaking, it not only optimizes the power grid operation, but also provides users with better power consumption services.

[0053] Example 2 See also Figure 2 As shown, the part not described in detail in this embodiment is described in Example 1, which provides an intelligent time-delay closing electric energy meter control method, including: Step 1: Obtain electricity consumption data and electricity consumption environment data of the electric energy meter; Step 2: construct a multidimensional data space based on the electricity consumption data and electricity consumption environment data of the energy meter; Step 3: Based on the multidimensional data space, a two-dimensional grid evolution model is constructed; Step 4, identifying the peak power consumption area in the two-dimensional grid evolution model; Step 5: For the identified peak power consumption area, the improved sphere optimization algorithm is used to determine the optimal delayed closing time and duration.

[0054] Example 3 This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the operation mode of the intelligent delayed closing electric energy meter control method provided above is implemented.

[0055] Since the electronic device introduced in this embodiment is an electronic device used to implement a smart time-delay closing electric energy meter control method in the embodiment of the present application, based on the smart time-delay closing electric energy meter control method introduced in the embodiment of the present application, the technical personnel of the field can understand the specific implementation of the electronic device of the present embodiment and its various variations, so how the electronic device implements the method in the embodiment of the present application is not described in detail here. As long as the technical personnel of the field implement the electronic device used in the smart time-delay closing electric energy meter control method in the embodiment of the present application, it belongs to the scope of protection of the present application.

[0056] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0057] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. An intelligent time-delay closing electric energy meter control platform, characterized in that: include: A data acquisition module is used to obtain the electricity consumption data and electricity consumption environment data of the electric energy meter; The space construction module constructs a multi-dimensional data space based on the electricity consumption data and electricity consumption environment data of the energy meter; The model building module builds a two-dimensional grid evolution model based on the multi-dimensional data space; An identification module is used to identify the peak power consumption area in the two-dimensional grid evolution model; The optimization control module is used to determine the optimal delay closing time and duration for the identified power consumption peak area by using an improved sphere optimization algorithm; each module is connected by wired and / or wireless means.

2. The intelligent time-delay closing electric energy meter control platform according to claim 1 is characterized in that: The electricity consumption data includes reading data, electricity power curve, electricity load type, electricity equipment type and user type; the electricity consumption environment data includes meteorological data, holiday information, electricity consumption area information and grid load level.

3. The intelligent time-delay closing electric energy meter control platform according to claim 2 is characterized in that: The method of constructing the multidimensional data space includes: The electricity consumption data and electricity consumption environment data of the electric energy meter are cleaned and standardized, and the categorical data are numerically encoded to obtain the preprocessed electricity consumption data and electricity consumption environment data; the preprocessed electricity consumption data and electricity consumption environment data are merged into a data set X; each row in the data set represents a sample; the sample is constructed into a vector representation, which is recorded as a sample vector; For each sample x in the data set X, calculate the distance between it and all other samples; based on the obtained distance, construct an n×n distance matrix D, where n is the number of samples in the data set X, and the element D(i, j) in the distance matrix D represents the distance between the i-th sample and the j-th sample; For each sample x, find the k samples closest to x in the distance matrix D, that is, the k nearest neighbors of x; k is a preset value and an integer greater than 1; the sample x and its k nearest neighbors form a set, recorded as the nearest neighbor set N(x); construct an undirected weighted graph G=(V, E, W), where V is the set of all nodes in the undirected weighted graph, corresponding to all samples in the data set X; E is the set of all edges in the undirected weighted graph; W is the set of edge weights; convert the undirected weighted graph G into a fuzzy geometric graph; optimize the fuzzy geometric graph for dimensionality reduction to obtain a multidimensional data space.

4. The intelligent time-delay closing electric energy meter control platform according to claim 3 is characterized in that: The method of finding the k samples closest to x in the distance matrix D includes: For each sample x, initialize k empty lists, copy the row corresponding to sample x in the distance matrix D to a temporary list T1, sort the temporary list T1 to obtain an ordered list in increasing distance, and divide the ordered list into k sublists of equal size; Take out the element with the smallest distance from each sublist to form a minimum heap, take out the element with the smallest distance from the minimum heap, add it to the ordered list S, insert the next element in the sublist where the taken element is located into the minimum heap, and repeat until all sublists are merged into the ordered list S. In the ordered list S, take out the first k elements, and their corresponding samples are the k nearest neighbors of x.

5. The intelligent time-delay closing electric energy meter control platform according to claim 4 is characterized in that: The method of converting the undirected weighted graph G into a fuzzy geometric graph includes: For each node in the undirected weighted graph G , calculate its comprehensive local density ; ； in, For nodes The nearest neighbor set The nodes inside For Node and nodes The distance between is a control parameter, and its value is in the range of (0, 1); for To its The distance to the nearest neighbor, For Node and nodes The importance factor between Importance Factor ;in, For Node and nodes The similarity score between For Node and nodes The connectivity score between is the weight coefficient, and its value is in the range of (0, 1); ;in, For Node The corresponding sample vector is For Node The corresponding sample vector is It is the vector dot product operation; If the node and nodes There exists a shortest path in an undirected weighted graph G ;but ; If the node and nodes In the undirected weighted graph G, they are directly connected. ; Find the ratio The comprehensive local density All nodes with large size form a set, which is recorded as the large density set ,for Each node in ,calculate and The distance between ,Pick All nodes in The minimum distance ; For each node , calculate a clustering factor , sort all nodes in descending order according to the value of the clustering factor, start from the node with the largest clustering factor value, divide it and its nearest neighbors into a cluster, then move to the next undivided node with the largest clustering factor value, and repeat until all nodes are divided into clusters; For each cluster, it is regarded as a simple geometry; for each node belonging to the cluster, its membership is defined as the inverse of the distance between the corresponding node and the central node of the corresponding cluster; the combination of the membership of all simple geometries and their corresponding nodes constitutes a fuzzy geometric graph Gg.

6. The intelligent time-delay closing electric energy meter control platform according to claim 5 is characterized in that: The method of performing the optimization and dimensionality reduction process includes: Define a low-dimensional space and determine the target dimension d of the low-dimensional space; for each sample x in the data set X, take it as a data point and project it into the d-dimensional space to obtain a d-dimensional vector as the initial position of the data point in the low-dimensional space; in the low-dimensional space, based on the initial position of the data point, construct a low-dimensional fuzzy simple geometric graph G_low; Define the loss function ; ； in x belongs to the convex hull shape in the fuzzy geometric graph Gg The membership degree of is the degree of membership of x in the low-dimensional fuzzy simple geometric graph G_low to the convex hull shape s; is the weight coefficient of the regularization term, is the neighbor distribution of x in the fuzzy geometric graph Gg, is the neighbor distribution of x in the low-dimensional fuzzy simple geometric graph G_low; Represents KL divergence; based on the loss function, the position of the data points in the low-dimensional space is iteratively optimized to minimize the loss function. In each iterative optimization, based on the position of the current data point, a low-dimensional fuzzy simple geometric graph is constructed, the value of the current loss function is calculated, and the position of each data point is back-propagated to obtain the gradient, and the optimizer is used to update the position of each data point according to the gradient; repeat until the preset maximum number of iterations is reached; the final low-dimensional fuzzy simple geometric graph is the multidimensional data space.

7. The intelligent time-delay closing electric energy meter control platform according to claim 6 is characterized in that: The construction method of the two-dimensional grid evolution model includes: The multidimensional data space is mapped onto a two-dimensional plane, and an N×N square grid is constructed on the mapped two-dimensional plane. Each grid point in the square grid represents a local area in the multidimensional data space. Each grid point is randomly assigned an initial state, which is either alive or dead. For each grid point, its neighbors are defined, and the neighbors include 8 adjacent grid points in the up, down, left, right, and four diagonal directions. For the square grid, set the evolution rule. The evolution rule is: if a grid point in the live state has 2 to 3 neighbors in the live state, it will remain in the live state in the next generation. If a grid point in the live state has less than 2 or more than 3 neighbors in the live state, it will become dead in the next generation. If a grid point in the dead state has exactly 3 neighbors in the live state, it will become alive in the next generation. Define the length of the time step; starting from the initial state, iteratively update the states of all grid points according to the evolution rules. In each iteration, for each grid point, find its neighbors in the grid and calculate their states. According to the evolution rules, determine the state of the corresponding grid point in the next time step, and update the states of all grid points at the same time; and record the number of grid points whose states are alive at each time step. The grid whose states are iteratively updated is the two-dimensional grid evolution model.

8. The intelligent time-delay closing electric energy meter control platform according to claim 7 is characterized in that: The method for identifying the peak power consumption area includes: Set a time window of fixed length, slide the time window along the time axis, and move one time step each time; in the process of sliding the time window, calculate the activity of each grid point in the time window; the calculation formula of activity is: ;in, is the number of time steps in the time window, is the index of the time step, Grid In time step status, For the time step Time Grid The number of live neighbors among the surrounding 8 neighbors, is the neighborhood influence coefficient, and its value is in the range of (0, 1); Grid activity level; is the step size control parameter; The activity of the grid points in the time window is organized into a matrix, recorded as the activity matrix, and Gaussian smoothing or mean filtering is applied to the activity matrix; a high activity threshold is defined, the activity matrix is ​​traversed, and the continuous area formed by the grid points whose activity is higher than the high activity threshold is marked as a potential peak area; Convert the activity matrix into a binary image, and set the pixels corresponding to the potential peak area on the binary image to 1, and the others to 0; obtain a preliminary binary image; apply dilation and erosion operations to the preliminary binary image; obtain the potential peak area in the middle section; In the process of sliding the time window, for adjacent time windows, calculate the overlap of the potential peak areas in the middle section obtained by them; define an overlap threshold; if the overlap between the current time window and a previous time window is higher than the overlap threshold, then record the duration between the two time windows; A minimum duration threshold mdu is set, and those mid-segment potential peak areas whose duration exceeds mdu are marked as stable areas. The tracking period and average threshold are set. For each pair of stable areas, their average overlap during the entire tracking period is calculated. If the average overlap is higher than the average threshold, the corresponding two stable areas are merged, that is, the union of the two corresponding stable areas in all time windows is calculated, and this union is used as the new merged area. The new merged area is morphologically processed to obtain the final power consumption peak area.

9. The intelligent time-delay closing electric energy meter control platform according to claim 8 is characterized in that: The optimal method for determining the delayed closing time and duration includes: Define decision variables, including the delay closing time and duration ; Define the objective function ; ； in , , as well as is the target weight coefficient, for The original power consumption at the time, To implement delayed closing The power consumption at the time, represents the positive part, is the cumulative effect factor of historical delayed impact; is the user impact factor, is the load balancing factor; ;in, is the power change rate weight coefficient; is the historical impact weight coefficient; for The power change at each moment, is the time interval; is a delayed time point in history; is the time decay constant; ;in, For the The importance weight coefficient of the class user, for Moment The number of users of the class; For the Class User The accumulated delay time of the moment; ;in, is the load balance factor; is the maximum load value during the observation period, To implement delayed closing The actual load value at the moment, is the minimum load value during the observation period; Randomly generate N2 spheres, where N2 is an integer greater than 1. Each sphere contains a center position and a radius. The center position is composed of decision variables, i.e. (x1, x2). All spheres are sorted in descending order according to the function value of the objective function. For each sphere, randomly generate M candidate solutions on the surface of the sphere, calculate the function value of the objective function of the candidate solution, take the candidate solution with the largest function value as the optimal candidate solution, update the center position of the sphere to the position of the optimal candidate solution, and adjust the radius of the sphere. The formula for adjusting the radius of the sphere is: ;in, is the adjusted radius, is the sphere corresponding to the optimal candidate solution The radius of The sphere selected from all spheres The sphere with the largest function value of the candidate solution ; is the random disturbance factor, and its value range is (-0.1, 0.1); All spheres are sorted, spheres with smaller function values ​​are eliminated, and spheres with adjusted radius are used as new spheres; the change rate of the optimal candidate solution is iteratively and repeatedly calculated. If the change rate is less than the preset change threshold for K consecutive times, the last optimal candidate solution is used as the final solution, and the decision variables corresponding to the final solution are extracted to obtain the optimal delayed closing time and duration.

10. A method for controlling an intelligent time-delay closing electric energy meter, which is implemented based on the intelligent time-delay closing electric energy meter control platform according to any one of claims 1 to 9, characterized in that: include: Step 1: Obtain electricity consumption data and electricity consumption environment data of the electric energy meter; Step 2: construct a multidimensional data space based on the electricity consumption data and electricity consumption environment data of the energy meter; Step 3: Based on the multidimensional data space, a two-dimensional grid evolution model is constructed; Step 4, identifying the peak power consumption area in the two-dimensional grid evolution model; Step 5: For the identified peak power consumption area, the improved sphere optimization algorithm is used to determine the optimal delayed closing time and duration.

Images